sebastianruder / NLP-progress

Repository to track the progress in Natural Language Processing (NLP), including the datasets and the current state-of-the-art for the most common NLP tasks.

https://nlpprogress.com/

MIT License

22.61k stars 3.62k forks source link

New task: noun compound interpretation #217

Open girishponkiya opened 5 years ago

girishponkiya commented 5 years ago

Noun compound interpretation

The semantic interpretation of noun compounds (NCs) deals with the detection and semantic classification of the relations between noun constituents.

Example: spoon handle => PART-WHOLE, student protest => AGENT, fee-hike protest => CAUSE.

Noun compound interpretation is a well-studied task. For more information, one may refer Girju et. al. (2005) and Nakov (2013).

Question:

Where should I add this?
Under an existing category? If yes, which one?

Notes:

This is NOT Semantic Role Labeling as the relations are not between verb and noun.
This is NOT Relationship Extraction as the relations are based on the context. Interpretation of noun compound can be done via paraphrasing (student protest => protest by student(s)) which is not case with relation extraction.

sebastianruder commented 5 years ago

How about adding this to shallow syntax? Would that work?

girishponkiya commented 5 years ago

No. This task is at the semantic level, and it is "mostly" out-of-context. I would suggest making a new category, named Noun Compound Interpretation.

sebastianruder commented 5 years ago

Good point. I'm a bit hesitant to create a separate category for it as it's quite a specialized task. How about we add it to relationship extraction for now but make it clear that this setting does not depend on the context?

girishponkiya commented 5 years ago

I agree that it looks tiny and a specialized task. But, the representation of relation makes it totally different. For instance, a semantic relation in a noun compound can be represented by means of..

prepositional paraphrasing: e.g., student protest => "protest by student", fee-hike protest => "protest against fee-hike" (Ponekiya et al., 2018)
fine-grained paraphrases using verbs: e.g., malaria mosquito can be paraphrased using carry, spread, cause, transmit, etc. (Nakov, 2008)

From this relation representation perspective, the task is totally different from Relationship Extraction.

Additionally, there are many different types of relation inventories and datasets, and proposed system for those datasets. Adding this task to Relationship Extraction will complicate the things.

tianjianjiang commented 5 years ago

If I may suggest a rough idea: how about using “multiword expression” as an umbrella category for the time being?

girishponkiya commented 5 years ago

@tianjianjiang, that's a good idea. We can have two different sections for conventional multiwords (mostly non-compositional) and noun compounds (mostly compositional).

sebastianruder commented 5 years ago

Thanks for adding your opinion, @tianjianjiang. I agree with the proposal. Feel free to create a Multiword expression category and add this task there. 👍